Uncertainty analysis of radar QPE in distributed hydrological modeling at Skjern River catchment, Denmark

Weather radar based Quantitative Precipitation Estimation (QPE) has been recognized as a reliable data input in distributed hydrological models due to its high spatial and temporal resolution. However since radar measures meteorological systems remotely and indirectly, the QPE product may contain uncertainties of many sources. The uncertainties in QPE will propagate through hydrological models and may result in amplified uncertainties in simulated hydrological responses.

The present study demonstrates a method for quantifying radar rainfall uncertainties from a statistical ensemble approach. The random sampling technique used to generate stochastic uncertainty fields is based on Sequential Gaussian Simulation (SGS). The hydrological impact of the uncertainties in radar QPE is analyzed by using the rainfall ensemble in a distributed hydrological model developed for Skjern River catchment in western Denmark.

The study shows that the uncertainties in radar QPE may be represented by 200 rainfall realizations. The simulated stream discharge is proportional to the intensity of the rainfall input signal. The coefficient of variation is calculated for simulated stream discharge and groundwater recharge at subcatchment scales. The results reveal strong scale dependency showing higher variations of hydrological uncertainties at smaller catchments.